It was an online assessment on Code Signal that was on a Jupyter Notebook that involved data cleaning and model construction, and model evaluation. This was following a resume submission.
Domande di colloquio [1]
Domanda 1
Build any prediction model given some data that needs to be cleaned.
Code test is not too hard but highly time-sensitive — like a mini data science project requiring quick EDA, modeling, prediction, and concise analysis. Accuracy, clarity, and efficiency matter most.
Ho presentato la mia candidatura online. La procedura ha richiesto 3 settimane. Ho sostenuto un colloquio presso Walleye Capital (Boston, MA) nel mese di ago 2025
Colloquio
Fairly standard 45 minute interview about a variety of statistical questions. We first spent 15 minutes where he asked me several questions about my research background. After that we jumped into several questions about measuring a models performance, trading strategies, and pure probability.
Domande di colloquio [1]
Domanda 1
If you have a skewed data distribution with only 10 true positives and 190 true negatives, how would you measure your predictive model's performance.